Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

How to create prior probabilities without using SAS EM

Reply
New Contributor
Posts: 3

How to create prior probabilities without using SAS EM

Hi everyone,

I would like to create prior probabilities exactly like we can do it in SAS EM. However, I would like to run everything in SAS Base, so I was wondering if I need to create a specific code for that?

Any ideas?

Many thanks!!

Super User
Posts: 17,750

Re: How to create prior probabilities without using SAS EM

StephanieB wrote:

"create prior probabilities exactly like we can do it in SAS EM"

Where in SAS EM are you creating prior probabilities that you'd like to replicate?  Logistic Regression or something else?

New Contributor
Posts: 3

Re: How to create prior probabilities without using SAS EM

Yes sorry I should have precised. I am using a logistic regression.

My training sample is 50/50, although my event is less than 2% in my whole population.

I would like to add that in SAS base.

In SAS EM, I just precise the prior probabilities by right clicking on the dataset >> Prior probabilities.

Super User
Posts: 17,750

Re: How to create prior probabilities without using SAS EM

You should check what the prior probability default is in EM, but I think what you're looking for is

1/p and 1/(1-p) where p is the probability of your event.

Please do double check, because it really depends on what your question is and other situations. It would help if you provided more details, and I also don't have time today to dig into this.

Respected Advisor
Posts: 2,655

Re: How to create prior probabilities without using SAS EM

I think where was going was that you really cannot calculate priors from logistic regression in Base SAS--you will need SAS/STAT to access one of the procs capable of logistic regression (LOGISTIC, GENMOD, GLIMMIX, CATMOD).

Steve Denham

Ask a Question
Discussion stats
  • 4 replies
  • 523 views
  • 0 likes
  • 3 in conversation